Great news Sergey! It definitely makes sense to create a CL if you've made a
good progress! Currently I have only turned length of a FixedArray to a smi
(I applied patch by Kevin Millikin and fixed several places appeared after
he did it), but I am still debugging some strange crashes.


--
Vyacheslav Egorov


On Fri, Apr 23, 2010 at 4:43 PM, Sergey Ryazanov <[email protected]> wrote:

> Yes. Right now I'm making the string length represented as a SMI. My goal
> is to use code generated in EmitNamedLoad for expressions like 'str.length'
> (and 'array.length' in the future).
>
> Vyacheslav, what stage is your work about heap object layout on? Does it
> make sense for me to prepare a separate CL for String::length?
>
>
> On Fri, Apr 23, 2010 at 5:16 PM, Anton Muhin <[email protected]> wrote:
>
>> Sergey, hear, hear :)
>>
>> I suspect you'll turn String and FixedArray lengths into Smis, correct?
>>
>> yours,
>> anton.
>>
>> On Fri, Apr 23, 2010 at 5:11 PM, Vyacheslav Egorov <[email protected]>
>> wrote:
>> > Anton,
>> >>> Any ETA for this?
>> > I think I should finish it by the end of the next week.
>> > Currently I am fighting with heap objects layout: I need to replace all
>> > integer fields in objects that might appear in oldspace with smi fields
>> to
>> > allow fast iteration without map decoding.
>> > --
>> > Vyacheslav Egorov
>> >
>> >
>> > On Fri, Apr 23, 2010 at 2:55 PM, Anton Muhin <[email protected]>
>> wrote:
>> >>
>> >> Vyacheslav,
>> >>
>> >> very good news.  Any ETA for this?  I'd rather postpone my change
>> >> until this new infrastructure is in to see if the trick would worth
>> >> it.
>> >>
>> >> yours,
>> >> anton.
>> >>
>> >> On Fri, Apr 23, 2010 at 4:51 PM,  <[email protected]> wrote:
>> >> > Anton,
>> >> >
>> >> > Just to let you know: I am currently cooking changes to replace
>> current
>> >> > rset
>> >> > writebarrier with card marking one.
>> >> >
>> >> > It seems that such tricks will either become unnecessary or will
>> become
>> >> > very
>> >> > simple, e.g. if we have a single dirty word per page [card = page] we
>> >> > can
>> >> > avoid
>> >> > writebarriers in swap completely.
>> >> >
>> >> >
>> >> > http://codereview.chromium.org/1737007/show
>> >> >
>> >
>> >
>>
>
>

-- 
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev

Reply via email to